Rapid Machine Tool Verification for Calibrating CNC Machines

Automated go / no-go checks improve quality, reduce rework and scrap.

State of the art rapid machine tool verification methods are fully automated and take just a few minutes. They are able to fully verify that a machine tool is operating within tolerance, before critical machining operations are carried out on high-value components.

Traditional methods of calibrating machine tools require considerable down-time as well as highly skilled labor. In the past, this has meant that machine tools were meticulously calibrated when they were built. A full recalibration would then only be performed if errors were observed in the parts they produced. In the pursuit of improved quality and zero-defects, many manufacturers now perform regular checks and recalibrations. Improved methods can reduce the time required for general health-checks to around 20 minutes and full calibrations to a few hours. This means that checks may be performed weekly and recalibrations annually. This is a major step forward, although a significant risk of non-conformance remains.

One approach to further reducing the risk of non-conformance is to integrate laser systems into machine tools in such a way that they can be fully self-calibrating. I recently wrote about the Etalon multiline system which can achieve this for certain types of machine. This integrates laboratory grade metrology into the machine, giving direct traceability to the frequency definition of the meter. In theory, this will enable a full calibration to be carried out in minutes with no human intervention. In practice, such an approach is only suitable for large machines and is very expensive.

An alternative approach is to perform a rapid verification test rather than a full calibration. A calibration will independently quantify each error source, allowing compensation of these errors. A verification test, on the other hand, may have sensitivity to all error sources, without being able to separate them. This means that regardless of the error source, the verification test will identify when there is an issue with the machine. It will not however, enable compensation for that error. Instead, once the problem has been detected, a calibration would then have to be carried out.

Rapid verification checks are complementary to the condition monitoring of machine tools which are being developed. It is anticipated that Industry 4.0 will see manual condition monitoring replaced with embedded sensors. For example, the cutting tool monitoring system I wrote about recently or the Cyber Physical systems for machine tools being developed by Sandvik to enhance quality and efficiency. Such an approach can completely eliminate the need to stop a machine to perform checks and monitor wear. However, it is not possible to detect all machine errors using embedded sensors, and therefore some down-time is still required to verify the accuracy of machine tools. This is especially important before removing metal from very high value components. Some aerospace components may already have a value of $100,000 when then arrive at a machining station. If a machine tool has developed a fault this could lead to the component being scrapped. Taking the machine off-line for a few minutes is therefore justified if it significantly reduces the risk of the component being scrapped. Rapid verification tests have, therefore, been developed which are able to check that machine tool accuracy is within specified limits, before cutting metal on such high value components.

Sources of Errors in Machine Tools

Machine tools produce inaccurate parts due to a number of error sources. The most commonly considered sources are kinematic errors. Most machine tools have a number of axes arranged in series. For example, a three axis milling machine has x, y and z axes. For a given commanded position along one of these axes, six positional errors are possible, corresponding to the 6 degrees of freedom governing any rigid body motion. For example, motion along the x-axis may have translational errors in x, due to the x-axis encoder, and in y and z, due to the straightness of the x-axis. Motion along the x-axis may also have rotational errors; rotation around the axis is often described as roll, while the two rotations about the perpendicular axes are described as pitch and yaw.

Any position within the volume of the machine is described by the position of each axis. For a three axis machine the nominal position is therefore given by three commanded coordinates. Since each axis has six degrees of freedom, the actual position is determined by 18 kinematic errors. Typically, the alignment or straightness between the axes is considered separately. It is therefore said that there are 21 kinematic errors for a three axis machine. However, the three straightness errors each have a single value for the machine. The other errors are dependent on the position along the axis and are therefore measured at a number of discrete positions with interpolation performed between these. For a typical machine approximately 200 individual correction values would be measured in a full calibration.

The conventional approach to kinematic errors, described above, assumes that each axis has errors which vary only with position along that axis, independent of the positions along other axes. This assumption generally produces a sufficiently accurate error correction model. There is however, some influence between axes, which means that a different approach, volumetric compensation, can produce a higher accuracy.

Other machine-tool errors include hysteresis effects such as backlash, loads such as cutting forces and workpiece, thermo-mechanical effects, dynamic forces, cutting tool wear, and motion control and software errors.It is important to note that most of these additional error sources will not be detected, or compensated by, a machine-tool calibration.

Physical Error Separation Methods

Physical error separation methods seek to isolate and measure individual errors one at a time. This approach is often used for kinematic errors, backlash, spindle errors and cutting tool wear. Traditional methods use levels, straight edges, dial indicators and slip gauges. The modern approach uses a laser interferometer. In both cases the instrumentation must be carefully aligned with the axis and different measurements performed for each degree of freedom along the axis. This is a time-consuming and highly skilled task but it enables all of the machine tools’ kinematic errors to be calibrated and corrected. The latest laser interferometer systems, such as Renishaw’s XM-60, use multiple laser beams to enable all six degrees of freedom to be measured with a single setup. This saves a considerable amount of time but still requires skilled manual alignment and significant machine downtime. The laser calibration can only calibrate kinematic errors and backlash. Separate calibrations are required for spindle errors and cutting tool wear.

Figure 1: Renishaw XL-80 Laser Interferometer setup to measure the positional error along the axis.

Figure 1: Renishaw XL-80 Laser Interferometer setup to measure the positional error along the axis.
Figure 2: API Spindle Analyser uses proximity sensors to measure the movement of a precision steel cylinder as it is spun by the spindle.

Figure 2: API Spindle Analyser uses proximity sensors to measure the movement of a precision steel cylinder as it is spun by the spindle.

Simultaneous Error Parameter Estimation

Simultaneous error parameter estimation methods make use of a large number of individual measurements simultaneously. Taken individually, each measurement would not be capable of indicating which error sources were responsible for observed positional errors. However, when all of the measurements are considered together with an error model for the machine, it is possible to simultaneously determine all of the error parameters in the model. This typically involves a least-squares minimization to find the error parameters which, when input to the error model, give the best fit solution to the observed positional errors. The Etalon multiline system which I recently wrote about is an example of simultaneous error parameter estimation. As with laser calibration, these methods normally only calibrate kinematic errors and backlash. Separate calibrations are required for spindle errors and cutting tool wear.

Methods to Determine Combined Error Effects

Methods which determine the combined effect of multiple error sources are not suitable for calibration. They do not quantify the individual error parameters required for error correction. However, they can be very efficient ways to evaluate the overall accuracy of a machine tool. Perhaps the most robust method is a cutting test in which a standard test piece is machined and then measured, typically using a coordinate measurement machine (CMM). A cutting test has sensitivity to all error sources; however, it is also very time consuming, both on the machine-tool and on the CMM.

Other methods which determine combined error effects include ballbars and various forms of artefact probing. A ballbar is essentially a linear transducer with the ball bearing at each end. Each ball is mounted in a magnetic cup, one clamped to the machine bed and the other to the machine’s spindle. The machine tool then moves its spindle in a nominally circular path, centered on the ball mounted to the machine bed. Any deviation from the circular path is measured as a change in distance between the balls. By plotting this displacement against the two axes in which the circular movement takes place, it is possible to identify many types of the error. A ballbar test is, therefore, an excellent health check although it still requires around 20 minutes with manual setup and interpretation of results.

Figure 3: Renishaw QC20 Ballbar Mounted on the Bed of a 3-axis Milling Machine.

Figure 3: Renishaw QC20 Ballbar Mounted on the Bed of a 3-axis Milling Machine.

Applicability of Methods to Rapid Machine-Tool Verification

Physical error separation methods, such as laser calibration, require too much machine downtime and manual intervention to be used for rapid verification. They are also not sensitive to all machine errors, requiring additional tests. Ballbar tests provide an excellent quick health check that again is not rapid or automated enough for verification tests carried out between machining operations.

Artefact probing can provide an ideal solution for rapid machine-tool verification. Most modern CNC machine tools are now equipped with probes which can be automatically loaded using the tool changer. Probing a pre-calibrated artefact can verify that kinematic error parameters are within tolerance and provide traceability to any subsequent probing operation. Subsequent probing of roughing cuts, or sacrificial cuts made for testing purposes, can then verify complete system performance of the machine-tool with sensitivity to all error sources, including spindle errors, tool wear, and controller errors.

Strategy for Rapid Machine Tool Verification

Rapid verification using artefact probing involves three stages:

  1. Baseline probing is performed a short time after the machine has been calibrated, when it is known that the kinematic errors are well within tolerance. The artefact and probing routine are exactly the same as will be used in stage two. The coordinates of the probe surfaces from the artefact are stored on the machine-tool controller as R-variables.
  2. Probe artefact before cutting to verify that kinematic errors are within tolerance and provide traceability for subsequent probing operations. Probed surface location and direction ensures sensitivity to all the kinematic errors of the machine-tool with sufficient spatial range and resolution. The actual design of artefact and consideration of probe positions depend on the size and configuration of machine-tool being verified.
  3. Probe cut surfaces to fully verify machine-tool accuracy. These surfaces may be roughing cuts or specially made sacrificial cuts. As with the probed points, they must cover the range of positions and directions so that the machine is properly verified. It may also be important to include representative features such as interpolated radii to verify the performance of the machine under circumstances where dynamic and controller errors may become significant.

For smaller machines a monolithic artefact may be used, typically with an automated pallet loading system which is used to load workpieces as well as the artefact. For larger machines it may be more practical to use a modular artefact with a number of sections mounted to the machine-tool structure; in such a case it is important that effects such as thermal deformation or impacts do not cause errors in the machine-tool structure to be mirrored in the artefact. I helped to develop these methods in a research collaboration between the University of Bath and Rolls-Royce. An artefact which was used in this project is shown below with arrows indicating probing locations and directions which give sensitivity to specific kinematic errors. It should be noted that many probe points provide sensitivity to more than one error source; this is the reason why error separation is not possible with this method.

Figure 4: Sensitivity to Kinematic Errors for Monolithic Artefact

Figure 4: Sensitivity to Kinematic Errors for Monolithic Artefact

If you would like to find out more about this method you can read the full research paper published by the University of Bath and Rolls-Royce.